AnandTech Storage Bench - The Destroyer

The Destroyer is an extremely long test replicating the access patterns of very IO-intensive desktop usage. A detailed breakdown can be found in this article. Like real-world usage, the drives do get the occasional break that allows for some background garbage collection and flushing caches, but those idle times are limited to 25ms so that it doesn't take all week to run the test. These AnandTech Storage Bench (ATSB) tests do not involve running the actual applications that generated the workloads, so the scores are relatively insensitive to changes in CPU performance and RAM from our new testbed, but the jump to a newer version of Windows and the newer storage drivers can have an impact.

We quantify performance on this test by reporting the drive's average data throughput, the average latency of the I/O operations, and the total energy used by the drive over the course of the test.

ATSB - The Destroyer (Data Rate)

The WD Black SN750 improves slightly over its predecessor on The Destroyer, setting a new record for the highest average data rate from a TLC-based SSD.

ATSB - The Destroyer (Average Latency)ATSB - The Destroyer (99th Percentile Latency)

The SN750 has slightly lower average latency on The Destroyer than its predecessor, but the Phison E12-based Corsair MP510 is in the lead among flash-based SSDs. The MP510 and both the current and previous WD Black are essentially tied for 99th percentile latency.

ATSB - The Destroyer (Average Read Latency)ATSB - The Destroyer (Average Write Latency)

The SN750 improves on both average read and write latency scores, but the Corsair MP510 has a much better average write latency on The Destroyer than the WD Black or even the Intel Optane 900P.

ATSB - The Destroyer (99th Percentile Read Latency)ATSB - The Destroyer (99th Percentile Write Latency)

The 99th percentile read and write latency scores for the SN750 are essentially unchanged from the previous WD Black, leaving it with one of the best QoS scores for reads but a score for writes that is worse than several of last year's TLC-based competitors.

ATSB - The Destroyer (Power)

The total energy consumption of the new WD Black over the course of The Destroyer is about 4% lower than its predecessor, further increasing WD's lead in NVMe power efficiency.

SLC Cache Sizes & SYSmark 2018 AnandTech Storage Bench - Heavy
Comments Locked

54 Comments

View All Comments

  • Oxford Guy - Monday, January 21, 2019 - link

    It's weird to review minor products and not review something major like the GTX 960.

    It may be explainable without conspiracy but it's still weird, in terms of priorities.
  • Alistair - Friday, January 18, 2019 - link

    Can't wait to read the SX8200 Pro review. I've bought a few, and I believe they are the best performing drives for the money right now. Excellent. I stopped buying Samsung after they started denying warranties in Canada (and they don't seem to want to fix that). Anyways the Samsung drives cost almost twice the SX8200 pro and perform exactly the same pretty much. Maybe the SX8200 Pro is even faster than the 970 EVO honestly.
  • gglaw - Friday, January 18, 2019 - link

    A few from decent reviewers are already up for this and the HP EX950 which are small upgrades over the incredibly well-rounded drives they replaced. The SX8200 (non-pro) and EX920 go on so many massive sales it is hard to beat them until the newer drives start dropping in price. The average home user would never know whether they had a 970 Pro, SX8200, SX8200 Pro, or EX950 running so major price differences would make the decision for me. After my initial experiences with the SX8200/EX920, these have been all I've stocked for close to a year. Made me almost regret my 970 Pro. They regularly go on sale for ~$75 for 500GB, and $135 for 1TB versions so my SSD adventures have become rather boring with no close 2nd place I would even consider buying.

    I'm likely going to get a SX8200 Pro just because I can't help myself with new versions of my favorite drives, but I'm already 99% positive I'll be in the same boat of not being able to tell any difference with the small upgrade. Then I'll have buyer's remorse again like the 970 Pro, knowing the 1TB version of the cheaper drive is barely more than the new 500GB one (SX8200 1TB $135, Pro 500GB $115).
  • ajp_anton - Friday, January 18, 2019 - link

    What's wrong with WD's idle power consumption, and am I right assuming that that makes it unsuitable for laptops (mobile ones, not gaming)?
  • Billy Tallis - Friday, January 18, 2019 - link

    We test SSDs on a desktop, and that means we need to jump through some hoops to get PCIe power management enabled. I've never encountered a desktop motherboard that even has PCIe ASPM enabled by default, and when you are lucky enough to get a BIOS option to turn it on, you can't trust that to take care of everything. Even with the OS set to override the motherboard's settings, not all drives are able to enter their deepest sleep state on our testbed.

    I view this situation as being similar to DEVSLEEP for SATA drives. It's pretty likely that a laptop which was designed to use M.2 PCIe storage will have all the right firmware bits enabled to use the deepest power saving modes, but they're normally not used (or usable) on a desktop and I don't currently have equipment that can work around that.
  • hnlog - Saturday, January 19, 2019 - link

    WD Black NVMe has problem on Linux with default parameter.
    Is it fixed on the new model?
    https://community.wd.com/t/linux-support-for-wd-bl...

    I think WD should test before shipping the former model.
  • Billy Tallis - Monday, January 21, 2019 - link

    Basically every NVMe SSD vendor has shipped something that turned out to have serious power management bugs, most often with APST and only on certain host systems. It's pretty clear that no vendors (SSD or motherboard) are thoroughly testing those features before shipping, and instead just make sure that it works with a small handful of Windows configurations. But even the Windows NVMe driver is a moving target and new builds have caused problems.

    It would probably help if the UNH-IOL NVMe Integrator's List testing included APST, but their current test plan only checks whether the drive can handle manually setting power states. And even if they were more thorough, only a few vendors put consumer drives through that certification.
  • FXi - Sunday, January 20, 2019 - link

    I wonder when we'll see the upper end of sizes in consumer drives jump to 4TB. Durability seems to be ready. Perhaps consumer need isn't quite there. But if controllers can handle it and layers exist for it to be built to that size in the M2 format, you'd think that's where they would go next since prices have come out of the stratosphere.
  • eastcoast_pete - Sunday, January 20, 2019 - link

    Is it just me, or do these comparisons make the HP 920 look quite good? Not in terms of top performance, but in performance/price. Has anybody here had any experiences with 920 drives?
  • piasabird - Sunday, January 20, 2019 - link

    Why not compare similar products together? Why is one drive a 2 TB drive? Since 2TB has more save locations it may naturally be faster due to drive space, cache size, energy usage, etc. Maybe Anandtech doesnt use samsung drives because Samsung will not donate the drives for free but other companies would give them free gear to test.

Log in

Don't have an account? Sign up now